New to networking - does my router support CAKE?

[capncybo]

Traditionally QoS, FlexQoS, and AdaptiveQoS isn’t supposed to necessarily tackle the issues of bufferbloat. But rather it’s focused on allowing all of your devices to share your internet connection fairly in some way or limit certain devices from using too much. I found FlexQoS and AdaptiveQoS to be unreliable at fairly share my vary limited internet connection among devices because some of those devices use https or CDN tunnels which is in some cases is not able to determine the application that’s using the data. CDN/VPN’s are troubling as streaming services like Netflix uses them and it blatantly ignores QoS unless you use Bandwidth Limiter on devices with static ip then the whole device is targeted and limited not just a application.

I just wanted to chime in & point out how FANTASTIC this explanation actually is... + certain applications do not properly adhere to the QoS standards to begin with.
And more & more content is becoming encrypted all the time & if the router can't easily determine the packet class the packets should default to the lowest classification.
Worse yet is when the routers have to decrypt data content themselves to check the contents on the fly... as that's more & more router CPU resources CHEWED-UP.

So the trick with QoS is basically to set limits which are low enough to prevent your router CPU from RAILING or maxing out.
When buffers get full & CPU cycles are maxed... LATENCY @ best... but, Gitches or Crashes @ Worst.

In most homes & many work environments... each client typically has more CPU power & an equivalent or faster network card than the router.
But these clients are gonna tell that router they want that DATA right NOW !!!

And then certain people get mad, rant & call the router a POS.
The funniest thing to me is... the internet has changed drastically over the years but somehow a:
ASUS RT-AC66U : FCC approval date: 30 May 2012 ; (Est.) release date: 10 June 2012
or...
ASUS RT-AC66U B1 : Date First Available: Nov. 17 2016
is expected to perform without even the slightest delay for a competitive on-line first person shooter experience.
+
To add insult to injury, despite tons of good advice in two different threads...
We never saw a direct ethernet connection bufferbloat test direct to one PC (Without using the router).

Isn't that where they should start?
If they don't have acceptable bufferbloat results like that... Then any additional effort is simply a waste of time & money.
But as Tech9 likes to point out...
Bufferbloat tests aren't typically equivalent to the real-world usage case.

 

[dtaht]

occasionally I feel the need to raise the amount of clue with laughter: https://blog.apnic.net/2020/01/22/bufferbloat-may-be-solved-but-its-not-over-yet/

 

[dtaht]

@360NoScoper I'm one of cake's authors. I'm aware of how bad the variable "sag" can be on crowded cable, lte, and wifi networks, and over here is an effort to dynamically adjust it. https://forum.openwrt.org/t/cake-w-adaptive-bandwidth/108848/

It needs testers, and a port of the core algorithms to merlin after it proves out.

There's been some difficult statements made on this thread, some correct some not, it's always my hope more would read papers like this:

https://arxiv.org/pdf/1804.07617.pdf

or rfc8290 (and it's cites), to gain a deeper understanding.

 

[DJones]

@360NoScoper I'm one of cake's authors. I'm aware of how bad the variable "sag" can be on crowded cable, lte, and wifi networks, and over here is an effort to dynamically adjust it. https://forum.openwrt.org/t/cake-w-adaptive-bandwidth/108848/

It needs testers, and a port of the core algorithms to merlin after it proves out.

There's been some difficult statements made on this thread, some correct some not, it's always my hope more would read papers like this:

https://arxiv.org/pdf/1804.07617.pdf

or rfc8290 (and it's cites), to gain a deeper understanding.

Thanks for sharing your insight. And I’m sure a port to Merlin will be quite welcomed here once it’s ready.

 

[dave14305]

It needs testers, and a port of the core algorithms to merlin after it proves out.

Lua will be a tough sell on Merlin (only available via Entware), so maybe @Lynx original shell script is more appropriate for Merlin. Wait and see what the final verdict is over there.

GitHub - sqm-autorate/sqm-autorate: Adjusts bandwidth for CAKE by measuring load and RTT times

Adjusts bandwidth for CAKE by measuring load and RTT times - sqm-autorate/sqm-autorate

github.com

GitHub - lynxthecat/cake-autorate: Eliminate the excess latency and jitter terrorizing your 4G, 5G, LTE, Starlink or other variable rate connection!

Eliminate the excess latency and jitter terrorizing your 4G, 5G, LTE, Starlink or other variable rate connection! - lynxthecat/cake-autorate

github.com

 

[Viktor Jaep]

Advanced search results | SmallNetBuilder Forums (snbforums.com)

Next level let-me-google-that-for-you right there!

 

[Viktor Jaep]

@360NoScoper ... I have a suggestion... One way around this is bumping up your internet connection speed to a higher bandwidth, and, on top of that, investing in a higher powered router... AC86U, or AX86U are decent candidates to start with. I have 0 problems with latency... nor 8 streaming/techy family members yelling at me because of internet issues (with 55 devices on the LAN), but am dealing with 99 real-world problems!

 

[DJones]

He

@360NoScoper ... I have a suggestion... One way around this is bumping up your internet connection speed to a higher bandwidth, and, on top of that, investing in a higher powered router... AC86U, or AX86U are decent candidates to start with. I have 0 problems with latency... nor 8 streaming/techy family members yelling at me because of internet issues (with 55 devices on the LAN), but am dealing with 99 real-world problems!

Mikrotik 10gbe sfp+ or a pfsense router, then for wifi a https://www.asus.com/Networking-IoT-Servers/WiFi-Routers/ASUS-Gaming-Routers/RT-AX89X/

 

[Viktor Jaep]

Mikrotik 10gbe sfp+ or a pfsense router, then for wifi a https://www.asus.com/Networking-IoT-Servers/WiFi-Routers/ASUS-Gaming-Routers/RT-AX89X/

Did that thing come off from an alien spaceship OR WHAT!

 

[capncybo]

@360NoScoper I'm one of cake's authors. I'm aware of how bad the variable "sag" can be on crowded cable, lte, and wifi networks, and over here is an effort to dynamically adjust it. https://forum.openwrt.org/t/cake-w-adaptive-bandwidth/108848/

It needs testers, and a port of the core algorithms to merlin after it proves out.

There's been some difficult statements made on this thread, some correct some not, it's always my hope more would read papers like this:

https://arxiv.org/pdf/1804.07617.pdf

or rfc8290 (and it's cites), to gain a deeper understanding.

Wow, I hope my previous attempts at QoS simplification were not too terrible, or hard to stomach.
I did get about halfway through the pdf but... to be brutally honest... It was pushing my intellectual limits & understandings.
However with the deliberate acronym name of CAKE & the target being optimizing Home Networks.
I think yourself & others involved are well aware the technology needs to work almost unassisted & with little required expertise.
So basically combining what is almost impossible... Max throughput with Min Latency/Delay.
Hence kinda like having your CAKE & Eating it too ;-)
You Clever Cats.
+ Thanks for the detailed CAKE Education, I'm saving it (to re-read & refer too)
Even if it did make smoke come outta my ears & I think I heard some grinding noises...
It's been a long time since school or myself working on an engineering project.
Whole lotta RUST, mind you isn't that a newish programming language?
Hard for an old dog to keep up with all this.

EDIT:
I re-read it, kinda forced my way through it... kept telling myself 100% comprehension was not a requirement...
OW ow ow my... struggling brain
But we are all VERY fortunate for all your CAKE contributions.
THX

 

[Lynx]

Lua will be a tough sell on Merlin (only available via Entware), so maybe @Lynx original shell script is more appropriate for Merlin. Wait and see what the final verdict is over there.

GitHub - sqm-autorate/sqm-autorate: Adjusts bandwidth for CAKE by measuring load and RTT times

Adjusts bandwidth for CAKE by measuring load and RTT times - sqm-autorate/sqm-autorate

github.com

GitHub - lynxthecat/cake-autorate: Eliminate the excess latency and jitter terrorizing your 4G, 5G, LTE, Starlink or other variable rate connection!

Eliminate the excess latency and jitter terrorizing your 4G, 5G, LTE, Starlink or other variable rate connection! - lynxthecat/cake-autorate

github.com

Thanks for the mention and link to my GitHub repository. The shell script in my main branch is nice and simple and generally works. I imagine it could be made to work in Merlin without much trouble?

With some guidance from @moeller0 of OpenWrt (also one of the CAKE contributors, and in any case certainly a QoS expert), I am presently working on a completely new implementation in bash, which I am pretty excited about. It dispenses with the notion of a tick rate and will react to ping results as they come in.

Once that is complete, tested and optimised, then I think it may offer a good candidate for adapting for use in Merlin, if there is sufficient interest amongst Merlin users in adaptive bandwidth adjustment for CAKE.